McKinsey & Company has started putting job candidates in front of a fake client — one that happens to be an AI. According to The Verge, applicants for roles at the consultancy are asked to work through a simulated client engagement in which an artificial intelligence plays the customer, testing how the candidate manages a demanding, occasionally unreliable counterpart in real time.
The exercise is a notable twist on the case interview, the decades-old format in which consultants are handed a business problem and judged on how they structure their thinking. Where a human interviewer can be charmed, deflected or read for cues, an AI client does not tire, does not soften and — depending on how it is prompted — can be confidently wrong. That last property may be the point: a consultant who cannot notice when a counterpart is asserting something unsupported is a liability to the firm that bills for the advice.
For candidates, the format raises the bar on a skill that has become explicit in white-collar hiring: not whether you can use AI, but whether you can keep your footing while it is in the room. It also raises familiar questions. Automated assessments have a documented history of producing inconsistent or noisy signals, and a simulated client is only as good as the scenario behind it — including the failure modes its designers chose to build in.
McKinsey's move fits a broader pattern in 2026, as consultancies, banks and law firms rebuild entry-level pipelines around the assumption that generative tools will do much of the drafting, summarising and first-pass analysis that juniors once did. What remains for humans is judgment under pressure — which is, conveniently, exactly what a sparring partner that never gets tired is well suited to probe.
